Group membership

• Belonging is a fundamental human need, not a taste (Baumeister and Leary 1995)
• Our large brains co-evolved with living in large groups (Dunbar 1993)
• A group means an ingroup (share, cooperate) and an outgroup (compete, beware)

An old species

• Homo sapiens is roughly 300,000 years old (Hublin et al. 2017)
• Modern behaviour built up gradually, with no sudden “revolution” (McBrearty and Brooks 2000)
• For almost all of that time: small, mobile bands of a few dozen adults (Hill et al. 2011)
• The same minds now live in democracies of millions

Invented borders

Once we sort a continuum into categories, we

• overestimate differences between categories
• underestimate differences within them

The colour example

• The spectrum is continuous
• The colour names, and the borders between them, are ours

Figure 1: The visible spectrum, roughly 400 to 700 nm. The light changes continuously, and every border between named colours is drawn by us. Source: Wikimedia Commons, public domain.

From categories to bias

• The same distortion, applied to people
• Them: all alike, and very unlike us
• Us: a collection of individuals (Quattrone and Jones 1980)

Tajfel and Turner

Social identity is the part of the self that comes from group membership (Tajfel and Turner 1986)

1. Categorisation: sort people, yourself included
2. Identification: the group becomes part of who you are
3. Comparison: judge us against them

Minimal groups

• Assign people to groups on a trivial basis: Klee vs Kandinsky, dot estimates
• No history, no contact, no self-interest at stake
• People still favour their own group, and even trade gains to widen the gap (Tajfel 1970; Tajfel et al. 1971)
• It works even when assignment is openly random (Billig and Tajfel 1973)

Two kinds of bias

• Ingroup favouritism: liking and helping us
• Outgroup derogation: disliking and hurting them
• Distinct, and the first does not require the second (Brewer 1999)

Party as identity

• If a party is a group like any other, partisans are an us
• Then SIT predicts affect, not only disagreement: warmth for our side, hostility for theirs

“the tendency of people identifying as [partisans] to view opposing partisans negatively and copartisans positively” (Iyengar and Westwood 2015, 691; Iyengar et al. 2019)

Introduction

• Last week’s debate asked whether policy views have moved apart
• ISL: that is the wrong test
• Look at how partisans feel about each other, as social distance (Bogardus 1947)

Theory

• Party identification is a social identity
• Identity implies warmth for us and coldness toward them
• Campaigns make the identity salient, and negative ads feed the dislike

Research design

• Thermometers: how warm do you feel toward each party? (ANES, 1978 to 2008)
• Social distance: traits of party supporters, a child marrying across party lines
• Comparison: United States vs United Kingdom, 1960 vs 2008 and 2010
• Explanations: policy attitudes, exposure to campaigns

Results: thermometers

• It is the out-party that has changed

Results: social distance

• Policy views explain little of this, campaigns a little

Conclusion

• Affect, not ideology: Americans dislike the other side more, without disagreeing much more
• Partisanship is an affective bond more than a set of views
• Risk: opponents’ decisions may stop counting as legitimate

Introduction

• ISL showed what partisans say
• Survey answers can be filtered or exaggerated
• Does animus show up where people cannot hide it, and in nonpolitical choices?

Theory

• Benchmark: race, the deepest American divide
• Social norms restrain racial bias, no norm restrains partisan bias
• Expect partisan bias to be automatic, as large as racial bias, and driven by dislike of the out-party

Research design

1. Implicit: reaction-time test of party and race associations
2. Scholarship: choose between two students, one cue for party or race, grades varied
3. Games: split $10 with a copartisan or opponent, same or other race
4. Games with a control: separate favour for us from hostility to them

Results: does the test work?

• The implicit measure sorts people the way their self-reports do

Results: saying vs reacting

• Explicit answers are more polarized than automatic reactions

Results: party vs race

• The party gap is larger than the race gap, even implicitly

Results: leaners

Figure 8: Mean net feeling thermometer, Republicans minus Democrats, with 95% confidence intervals, by party identity including leaners and pure independents. Leaners sit close to the partisans, pure independents near zero. Source: Iyengar and Westwood (2015), Figure 5.

• Leaners behave like partisans, only pure independents are neutral

Results: scholarship

• Four in five partisans pick their own side

Results: merit and party

Figure 10: Predicted probability of picking the Republican, by participant’s party, for equally qualified candidates and for each candidate being more qualified. Partisan choices barely respond to grades. Source: Iyengar and Westwood (2015), Figure 6.

• Grades barely move partisan choices

Results: merit and race

Figure 11: Predicted probability of picking the African American, by participant’s race, for equally qualified candidates and for each candidate being more qualified. Racial choices respond to grades. Source: Iyengar and Westwood (2015), Figure 7.

• Party trumps merit, race does not

Results: games

• With real money at stake, party matters and race does not

Results: by party

• The bias is symmetric: each side gives its own about 68 cents more in the dictator game

Results: love or hate?

• In the trust game the penalty for opponents is about twice the bonus for copartisans

Conclusion

• The party divide is as deep as race, or deeper, and it reaches nonpolitical choices
• It is permitted: no norm restrains it
• Hostility toward them outweighs favour for us
• Elites have an incentive to fight rather than cooperate

Recap: takeaways

• Groups are old, and categories distort how we see people
• Social identity theory: a category alone is enough for bias
• Affective polarization: dislike of the other side, measured apart from policy views
• In the US it has grown since the 1980s, while in-party warmth stayed flat
• It shows up implicitly and in behaviour, and exceeds racial bias

Recap: learning goals

Goal 1. Favouritism and hostility from social identity theory

• Categorisation, identification, comparison, and the wish for positive distinctiveness
• Favouritism and derogation are distinct. Party seems to trigger both.

Goal 2. What affect adds to ideology

• Division can grow in feelings while policy views stay put
• ISL: out-party ratings fall, policy attitudes explain little

Goal 3. Animus beyond survey answers

• Implicit tests and costly choices show the same bias, with no norm to restrain it